TY - JOUR

T1 - Tensor relational algebra for distributed machine learning system design

AU - Yuan, Binhang

AU - Jankov, Dimitrije

AU - Zou, Jia

AU - Tang, Yuxin

AU - Bourgeois, Daniel

AU - Jermaine, Chris

N1 - Funding Information:
This work was supported by an NIH CTSA, award no. UL1TR003167 and by the NSF under grant nos. 1918651, 1910803, 2008240 and 1842494. We also thank the anonymous reviewers for their insightful comments on earlier versions of the paper.
Publisher Copyright:
© 2021, VLDB Endowment. All rights reserved.

PY - 2021

Y1 - 2021

N2 - We consider the question: what is the abstraction that should be implemented by the computational engine of a machine learning system? Current machine learning systems typically push whole tensors through a series of compute kernels such as matrix multiplications or activation functions, where each kernel runs on an AI accelerator (ASIC) such as a GPU. This implementation abstraction provides little built-in support for ML systems to scale past a single machine, or for handling large models with matrices or tensors that do not easily fit into the RAM of an ASIC. In this paper, we present an alternative implementation abstraction called the tensor relational algebra (TRA). The TRA is a set-based algebra based on the relational algebra. Expressions in the TRA operate over binary tensor relations, where keys are multi-dimensional arrays and values are tensors. The TRA is easily executed with high efficiency in a parallel or distributed environment, and amenable to automatic optimization. Our empirical study shows that the optimized TRA-based back-end can significantly outperform alternatives for running ML workflows in distributed clusters.

AB - We consider the question: what is the abstraction that should be implemented by the computational engine of a machine learning system? Current machine learning systems typically push whole tensors through a series of compute kernels such as matrix multiplications or activation functions, where each kernel runs on an AI accelerator (ASIC) such as a GPU. This implementation abstraction provides little built-in support for ML systems to scale past a single machine, or for handling large models with matrices or tensors that do not easily fit into the RAM of an ASIC. In this paper, we present an alternative implementation abstraction called the tensor relational algebra (TRA). The TRA is a set-based algebra based on the relational algebra. Expressions in the TRA operate over binary tensor relations, where keys are multi-dimensional arrays and values are tensors. The TRA is easily executed with high efficiency in a parallel or distributed environment, and amenable to automatic optimization. Our empirical study shows that the optimized TRA-based back-end can significantly outperform alternatives for running ML workflows in distributed clusters.

UR - http://www.scopus.com/inward/record.url?scp=85115288826&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85115288826&partnerID=8YFLogxK

U2 - 10.14778/3457390.3457399

DO - 10.14778/3457390.3457399

M3 - Conference article

AN - SCOPUS:85115288826

VL - 14

SP - 1338

EP - 1350

JO - Proceedings of the VLDB Endowment

JF - Proceedings of the VLDB Endowment

SN - 2150-8097

IS - 8

T2 - 47th International Conference on Very Large Data Bases, VLDB 2021

Y2 - 16 August 2021 through 20 August 2021

ER -